[Q] USB is very slow. Is this a known bug?

Naren (Narendra) Sankar nsankar at broadcom.com
Sun Nov 4 09:06:56 EST 2012

Unfortunately specs belong to synopsys and that is the constraint for us to release it. I don't believe that they have released the spec publicly and we can try to ask again. 
Naren Sankar
Broadcom Corporation
+886 975 355 267

----- Original Message -----
From: Greg KH [mailto:greg at kroah.com]
Sent: Friday, November 02, 2012 08:21 AM
To: Ezequiel Garcia <elezegarcia at gmail.com>
Cc: linux-rpi-kernel at lists.infradead.org <linux-rpi-kernel at lists.infradead.org>
Subject: Re: [Q] USB is very slow. Is this a known bug?

On Fri, Nov 02, 2012 at 12:00:32PM -0300, Ezequiel Garcia wrote:
> On Sat, Sep 22, 2012 at 5:37 PM, Greg KH <greg at kroah.com> wrote:
> > On Sat, Sep 22, 2012 at 05:25:33PM -0300, Ezequiel Garcia wrote:
> >> > Yes, the hardware is broken, the drivers are broken, and I would not
> >> > recommend anyone wanting to do any real USB data transfers to use this
> >> > platform at all.  Seriously.  Go complain to Broadcom if you want to see
> >> > this fixed, they are the ones with the specs that we can not get access
> >> > to.
> >> >
> >>
> >> Mmmm, I see.
> >>
> >> For some reason (cheapness? crazyness?)
> >> there are a few users out there wanting to do video streaming
> >> with their rpi and they want to use stk1160 devices.
> >> Thus my question.
> >
> > The answer is, "It will not work, buy a beaglebone or something real if
> > you wish to do that."
> >
> >> I don't want to bother you asking: what do you mean by "hw is broken",
> >
> > Just that, the USB controller in this board can not handle this type of
> > throughput.  If anyone doubts me, please show me the chip specs that
> > show me otherwise (yes, this is a ploy to get the specs from Broadcom,
> > but nothing else seems to work...)
> >
> >> "drivers are broken"?
> >
> > The USB driver code is horrible, really really horrible.  Seriously,
> > it's so bad that no one will touch it, myself included.
> >
> I've started looking into that usb code. I agree it's not the
> handsomest driver around...
> Could point me a usb host driver **similar** enough to start a
> massive refactoring effort?

Rumor has it that the Samsung submitted driver for the same core can be
"tweaked" to work with the rpi hardware, but, without the specs for the
hardware, verifying this is almost impossible.  I would work to get
access to those specs first, before wasting your time on this.

Good luck,

greg k-h

linux-rpi-kernel mailing list
linux-rpi-kernel at lists.infradead.org

More information about the linux-rpi-kernel mailing list